Jason Lemkin
Everything Jason personally uses, recommends, or has created — plus things they don't recommend — sourced from their own show and appearances on other podcasts.
Top picks
Ranked by repeated personal use, excluding sponsorships.
Personal use in 11 episodes
"Haiku 5.5... it's almost as good for a lot of use cases as Sonnet. And it's a tenth or a twentieth the price. It's shocking, it's so cheap. Haiku 5.5 is like the first reasoning model, and you can dial up how much reasoning you want, where you can almost not care about what it costs... We haven't had a reasoning-level model where you didn't have to stress every token, every dollar."
▶ 50:54"When I switched it myself to Haiku 5.5 and Jev, total costs went down about 70%, 72%... It's faster and 72% cheaper and better. Nothing got cheaper until Haiku 5.5 plus Jev. And I saw a 72% decline for B2B workflow running constantly, for the same task and higher quality."
"I ran it on Sonnet 5.5, which just came out. Input tokens 42% higher than before, output tokens 44% higher. It passed more of the blind test, quality went up. Cost went down about 10%."
"The other day I went into Claude and I just turned on the Google Drive connector — it's one of the three primary connectors. This is not esoteric, this is not third-party. It went in, scanned all my docs, found Jason's gems, found the ideas. The agent then went and changed my code and my algorithm without telling me."
"I'm in Fable. I've now moved to Opus 5."
"It uses Sonnet and Opus, but so there's different definitions of model. And it was good, but man, once I started labeling all of this, it got exponentially better."
"I couldn't solve this big algo problem — I solved it in about 20 minutes in Claude Opus. By using Claude plus Opus, I was able to get to the heart of the problem in an algorithm I could not understand. I lost a day using the n minus one step-down model. It was 20 minutes and actually zero because I get it in my $200 max account."
"This is running on Sonnet. Sonnet rapidly goal seeks. It tries not to finish complicated behaviors. I need to work with the agent to change how we do it and make clear all contracts must be read from beginning to end."
"Claude tells me this the other day. I was talking about my portfolio. He's like, 'Well, worst case, this one will exit for one and a half to two billion to one of these three folks.' I'm like, 'Thanks, Claude. I mean, I feel better.'"
"now that we've moved to Claude, she hasn't used OpenAI in four months"
"Claude, Sonnet, and Opus since 45 and 46 are so good. I want to stick there. It's so good"
Personal use in 8 episodes
"The biggest change for our little tiny team of two and a half humans is we run Replit between me and Amelia. We run it 20 hours a day now. When we started the show, it didn't quite work. At the end of last year, the models got better, and it'd be like an hour a day. Now, it's 10 to 12 hours a day. It is our partner, as our team."
▶ 19:14"I mean, we do use Replit about 14 to 15 hours a day. Basically, Replit has its own router in it, right? Its own model selector. It can be invisible or visible to you depending on what category you're in."
"I got switched over on Replit to auto router where it switches between Claude, Fable, the open source ones, and everything. Now I don't even know which one it's using. And then I saw some performance degradation. So then I had to switch it all back to Claude."
"I'm using the mix of the models in Replit — it's Sonnet plus open source basically what I'm using. After spending about 10 hours in Replit I couldn't solve this big algo problem."
"Replet, it does it automatically now. Replet if you have a complex feature replet builds it in cloud sonnet not open it builds it in sonnet to save money"
"I built a for fun yesterday on Replit, I built a fully functional website in about six minutes that has video, audio, and everything"
"I've basically used in Replit's very very good"
"I've been on Replit since June. And when I started, you couldn't finish anything... by December, especially with Opus45, all of a sudden it could finish apps that worked"
Personal use in 3 episodes
"We run Salesforce entirely headless. We have our own agent, our own AI VP of revenue. It runs Salesforce under the hood. Pro is it makes Salesforce much more powerful than it ever was. It's like I didn't log into Salesforce for seven years. Now I log in every day because I have an agent."
▶ 57:58"We run Salesforce headless. We don't even log into Salesforce — and these are opportunities and threats."
"We moved to headless Salesforce. Okay, we don't log into Salesforce anymore. We could swap that out for our own database. It's not worth it. It's not worth maintaining that database."
Personal use in 2 episodes
"I love open router. I talked about it on the show like six months ago. I'm a customer. I'm a user. It was brilliant. It was one of these pieces of software like ElevenLabs which is just instantly easier to deploy. It's just elegant. It was just a beautiful piece of software."
▶ 20:58"I love Open Router. I mean, I use it. It's just very interesting. It's a very simple way to dynamically pick which LLM to use"
Personal use in 2 episodes
"Higgsfield, where Harry and I are both investors. I was one of the first 10 users or customers. Higgsfield just announced they're at 500 million in revenue, actually doing 2 million a day now just in credit card billings outside of the enterprise. You can run them all on Higgsfield — Kling, Sora, Veo — you can just run them all in videos."
▶ 50:53"Higgsfield — I was one of the first 10 customers. I think what could you do? Make a 4-second video using Kimi or Qwen. Now you can make a full-length motion picture, and you can do it in another way. It's just so hard now — and that on Lovable, Replet — the moats are starting to get thick."
Personal use in 2 episodes
"I'm actually willing to make that bet, because I'm using Jev a lot and I can already see that if it never changed, maybe I would swap it out, right? But if it gets better than the rest, there's no way I would change... Jev is super fast and super cheap. And it is accurate enough to use for maybe 10% of your AI use cases. If they make it hyper accurate, it changes the game."
▶ 30:37"I've used it. It's awesome. What I use it for is deciding who in the SaaStr community should meet each other. It's quite good at that. Infinitely so cheap it doesn't even measure it. Ten times faster is the use case, but only for these limited use cases."
Personal use in 1 episode
"In the Bay Area, I got rid of my car. So, I only do Waymo and autonomous driving. I don't drive. I'm done with it. I would certainly never go back. Never go back to driving a car. It's just archaic."
▶ 44:33Warned against
"I think it's just Miro at scale, but with so much scale that it survives. It's a reminder that Miro didn't — in today's world, $600 million in collaboration would be scale, but Adobe has scale. Even if Adobe isn't really making the right moves in AI and it can't really afford it and it's just moving the deck chairs around for its C-suite... I mean, it's pretty bad. It's a nothing burger. It's a sign that nothing's going to change."
"I finally turned off Canva after 12 years. Don't need it... The agents don't need it."
"OpenAI and Anthropic have the two worst models that exist. They're called OpenAI Mini and Haiku. They're terrible. At least my little evals, they fail 100% of the time."
"OpenAI and Anthropic have the two worst models that exist. They're called OpenAI Mini and Haiku. They're terrible. Just go on Claude and switch to Haiku if you can and ask it a question — it doesn't remember anything, it's 2023, and it doesn't know where anything is."
"Fable then went and changed my code and my algorithm without telling me. Never got a notice. Never was told. Never was in a change log. Never was anywhere. I only found out later when the agent flashed 'conflict with Jason's Gems' when I was trying to fix something else."
"Figma make is one of the worst products I've used in the last six months"
"HubSpot today is sitting at 10 billion. The tough hand HubSpot has is it spent the last five years beating Salesforce at the low end. It's a CRM company now. It's not a marketing automation company — and now the low end is so good with AI."
"Loom is not the breakout success of Atlassian. Getting rid of collaborative free seats — which is how we all grew up using Zoom — that's a sign of just how hard it is, even if you beat the quarter. Whenever I see the base getting overly monetized or harvested, it's a sign of stress in the organization."
"the oldest piece of software uses Marketo. Okay, terrible email marketing service. About two weeks ago, it started to violate the CAN-SPAM act"
"I've paid for Netflix. I'm embarrassed to say how many years I've paid for Netflix... I've stealth turned off Netflix. I'm still paying. I may pay for another year, but not"
"Fable grabbed the draft notes out of hundreds of files in my drive, MCP'd into Replit on its own, and changed the core algorithm without telling me."
Recent episodes
All →Showing 137 of 164 products
Clear filtersMentioned in 7 episodes
"Grok is great for images."
"Grok Bot's an interesting one because SpaceX is public. One of the reasons they work is because they can break the rules."
"it appears that Cursor may be at least notionally the largest customer for Grok for coding"
"everything from already integrating Grock to data centers in space to crossing a trillion dollars in bookings"
"Why? Because Grok isn't good enough."
"These evals don't talk enough about speed. Like, this is why a lot of these things are useless, like Grok and others, because they're too slow or they time out."
"These aren't a set of issues that have been solved in the last year. They weren't solved with OpenClaw leaking everybody's confidential information. Now we have better guardrails, better harnesses. It's not solved with Grockbot, which looks like it's leaking everybody's information."
Mentioned in 11 episodes
"Cursor barely pulled it out. The Cursor V2 would have died, right? Reselling Anthropic tokens."
"There's some deep irony that a $60 billion core acquisition's running on China open-source models primarily. This is not going to last, right?"
"Everyone is building 10, 12, 14 hours a day in Claude Code or Cursor, whatever."
"Cursor has been swooped off the table. It was a standalone, more individual project. You really only have these guys and Cognition at this point who are basically going to the enterprise."
"Grok Bot was built in 5 weeks by the Cursor team or whatever."
"Cursor was super easy. It rockets to half a billion in revenue. Claude Code comes out and all it seems like everybody's moved — Cursor's dead, right? Cursor goes multimodel really early. It changes everything and rockets to a 60 billion outcome."
"Probably why Cursor wasn't dumb to sell at 60 billion. I think we can agree that's true."
"That led to the early folks that got whiplash, like Cursor having to go open source really early. And that's an interesting niche issue on X."
"the $60 billion cursor acquisition"
"barely anyone's using cursor today in in my own portfolio and people said it on Twitter too, like cursor is dead"
"everyone has moved off cursor and moved to Claude code"
Mentioned in 9 episodes
"Obviously some of this is because ChatGPT is so good for images, right?"
"Then there's like price performance and speed, and then I think they picked whatever GPT-5.5 medium or something, and they said the rest isn't worth it for Rippling today."
"Every ChatGPT release, every Higgsfield release, every L's, you can do more and more of its functionality in core AI. This is what we fear as investors — that you can do our investments in ChatGPT or Claude. That's the ultimate fear."
"There's 10 companies in middle America that are like, 'We have a ChatGPT subscription. Maybe next year we'll try some of this agentic stuff.'"
"You need your own micro model whether it is some sort of reasoning layer that you build on top of Claude or ChatGPT or Kimi."
"When you're in China, you can't access it. I just couldn't use ChatGPT or Claude or the APIs."
"The job today is not to screw around on ChatGPT and to be a prompt engineer"
"They refused to release Claude before ChatGPT because he didn't think it was safe"
"I'm one of the seven people that uses Claude instead of ChatGPT as a consumer app."
Mentioned in 6 episodes
"The demand is there. If you look at OpenRouter and others, the demand does not appear to have declined."
"We may see these open weight models go from 60-something percent of OpenRouter to 10% in 6 months and that would not be an extraordinary amount of change based on what we've seen."
"We're all open routing one way or the other... Unlike, you know, the average idiot configuring OpenRouter themselves hoping they get it right."
"Did you see the Cursor deal? Did you see Open Router? Open Router — we need more money, guys."
"I've talked about other routers. You can use 78 models if you want."
"In theory with Open Router and others you could route to 10,000 models. In practice, is the number three thing you're going to figure out what's the best open-source model I can use for my application?"
Mentioned in 1 episode
"It's great that a lot of stuff runs on WordPress or that Replit and Lovable and others immediately ran like gateways or barriers to protect you."
Mentioned in 5 episodes
"It's great that a lot of stuff runs on WordPress or that Replit and Lovable and others immediately ran like gateways or barriers to protect you."
"Certainly Replit, founded in 2016, is like a big case study where you just latch on to that and it just works."
"Replit's an example. It was freaking in the wilderness for 6 years until it added the models, right? It was a super nerdy web IDE. I think to answer Harry's question, it's a tough question to address as investors and employees."
"the Replit round that blows my mind when I think Craft and some other folks did it at a billion in Replit and like 21 preai"
"I was at Replit the other day at 300 million in revenue, 300 people. It's probably like the old days at Atlassian, right?"
Mentioned in 11 episodes
"It's great that a lot of stuff runs on WordPress or that Replit and Lovable and others immediately ran like gateways or barriers to protect you."
"It's just if you're Replit or Lovable or Base44 or v0, you have these extra incentives because your average user just wants to have a high quality output."
"One of the things that's under discussed that actually a Replit or Lovable did well in the early days is they could be on all the time."
"I do know that like the Replets, Lovables, Versels, and others, it costs about 3 to 4 bucks per person to deliver that."
"This will seem cheap when Replit's at 15 billion, Lovable's at 100 billion — everyone's out there because Linear will be the one powering them all."
"Lovable and Replet which — even though I'm a user for a year — you could have made fun of these products when Harry invested and when we started the show. Now they really can build almost production-grade, highly secure apps with everything across. It's just so — like I know we're building so quickly."
"I firmly believe a big chunk of Replit and Lovable could have been Figma's if they'd done it."
"Replit lovable vzero they're all good at they all pass the test"
"I tried lovable last week. The answer in 2026 is what commercial AI tool have you brought into your organization this month"
"you'd ask repler lovable to build an app... she built an entire AIVP of marketing that is better than anything on the market... Their website was all built in lovable"
"if you're on replet or lovable or Vercel or something, it will actually do penetration testing"
Mentioned in 1 episode
"So, like, this is like Mr. Robot live. Like, all these old websites are running on PHP. No one's patched them since this happened."
Mentioned in 7 episodes
"Why do you want to have this meeting? Let's just use Haiku 55. It's pretty good."
"He said, 'We're going from Claude to OpenAI because it's just as good and materially cheaper.'"
"Claude was just the same proof of concept app, but it worked much better for Anthropic."
"Rippling posted their view as a B2B player of what models they use. They said, in the world across Rippling, we looked at two things that were best for us: Opus 4.8 — it's an N-minus-one model but it's well trained with their harness."
"You can do our investments in ChatGPT or Claude. That's the ultimate fear."
"It's basically the ability to have Claude as a fully present member of a Slack channel focused just on whatever that is — be it your legal team, it's a Claude agent that's legal, that's just focused on legal, that just has access to that sort of information, but is a fully present member of your channel."
"They refused to release Claude before ChatGPT because he didn't think it was safe"
Mentioned in 1 episode
"Have you seen Office Space? Like it's not more humane to meet with someone in HR and get fired."
Mentioned in 1 episode
"Cohere is now a fourth-tier LLM, right? Fourth-tier player that has only sovereignty on its side as near as I can tell."
Mentioned in 1 episode
"I'm comfortable with the Fireworks at 40. This one at 30 just makes me nervous. Fireworks has met my 100-day rule. It's got 100 days of compounding genuine growth."
Mentioned in 2 episodes
"People throw out Zapier as the one-and-done example... really one venture round and they're at 300 million in revenue today... a great company that did something generational."
"Let's give it a week, right? Or is this Zapier on steroids?"
Mentioned in 1 episode
"I think Bending Spoons is like the Berkshire version of software, right? It's like See's Candy or Coke... Vimeo, kind of a dog that they kept. Evernote, kind of a dog. Now they made them work with price increases."
Mentioned in 4 episodes
"Everyone is building 10, 12, 14 hours a day in Claude Code or Cursor, whatever, but the amount of time it takes to review and process that to make it production intent is still an order of magnitude more than building."
"It's pretty annoying, you can't really keep Claude Code or Codex going 24 hours. That's why OpenClaw took off and others."
"You need to run this thing 8 hours a day like Claude Code or Codex to really matter."
"I don't even remember if Claude Code existed when we started the show or not. It might not have — it sounds so crazy. I'm not sure it had launched. Cloud Code comes out and all it seems like everybody's moved."
Mentioned in 1 episode
"If Kimi actually was so great, you know what? All those guys, Replit would be using it entirely... Amjad didn't get up on stage and say, 'We're routing everyone through Kimi,' did he?"
Mentioned in 1 episode
"I think Rory's in CodeRabbit, or I could be wrong, they're an epic player, right? There's all these great folks."
Mentioned in 2 episodes
"Claude Design or whatever the heck it's called, right? Which didn't kill Figma."
"We have no evidence design is going to kill Figma in any way, shape, or form at this point."
Mentioned in 1 episode
"Salesforce announced they've got their own models that are built on Nemotron and built with Nvidia."
Mentioned in 1 episode
"Or even Fin.ai, like, okay, I don't think they count. TechWolf and Fin.ai don't count."
26 recommended · 0 sponsored · 11 warned against · 0 episodes